Alzheimer’s disease (AD) is one of the most common neurodegeneration diseases caused by multiple factors. The mechanistic insight of AD remains limited. To disclose molecular mechanisms of AD, many studies have been proposed from transcriptome analyses. However, no analysis across multiple levels of transcription has been conducted to discover co-expression networks of AD. We performed gene-level and isoform-level analyses of RNA sequencing (RNA-seq) data from 544 brain tissues of AD patients, mild cognitive impaired (MCI) patients, and healthy controls. Gene and isoform levels of co-expression modules were constructed by RNA-seq data. The associations of modules with AD were evaluated by integrating cognitive scores of patients, Genome-wide association studies (GWAS), alternative splicing analysis, and dementia-related genes expressed in brain tissues. Totally, 29 co-expression modules were found with expressions significantly correlated with the cognitive scores. Among them, two isoform modules were enriched with AD-associated SNPs and genes whose mRNA splicing displayed significant alteration in relation to AD disease. These two modules were further found enriched with dementia-related genes expressed in four brain regions of 125 AD patients. Analyzing expressions of these two modules revealed expressions of 39 isoforms (corresponding to 35 genes) significantly correlated with cognitive scores of AD patients, in which 38 isoforms were significantly up-regulated in AD patients comparing to controls, and 33 isoforms (corresponding to 29 genes) were not reported as AD-related previously. Employing the co-expression modules and the drug-induced gene expression data from Connectivity Map (CMAP), 12 drugs were predicted as significant in restoring the gene expression of AD patients towards health, which include nine drugs reported for relieving AD. In comparison, four of the top 12 significant drugs were known for relieving AD if the drug prediction was performed by the genes expressed significantly different in AD and healthy controls. Analysis of multiple levels of the transcriptomic organization is useful in suggesting AD-related co-expression networks and discovering drugs.

阿尔茨海默氏病（AD）是由多种因素引起的最常见的神经退行性疾病之一。AD的机理见解仍然有限。为了揭示AD的分子机制，已经从转录组分析中提出了许多研究。但是，尚未进行跨多种转录水平的分析来发现AD的共表达网络。我们对来自AD患者，轻度认知障碍（MCI）患者和健康对照者的544个脑组织的RNA测序（RNA-seq）数据进行了基因水平和同工型水平分析。通过RNA-seq数据构建共表达模块的基因和同工型水平。通过整合患者的认知评分，全基因组关联研究（GWAS），替代剪接分析，和痴呆相关基因在脑组织中表达。总共发现29个共表达模块，其表达与认知评分显着相关。其中，两个亚型模块富含AD相关的SNP和其mRNA剪接显示出与AD疾病有关的显着改变的基因。进一步发现这两个模块富含在125位AD患者的四个脑区域中表达的痴呆相关基因。分析这两个模块的表达后发现39种亚型（对应35个基因）的表达与AD患者的认知评分显着相关，其中AD患者中38种亚型与对照相比显着上调，而33种亚型（对应29个基因） ）之前未报告为与AD有关。利用连接图谱（CMAP）的共表达模块和药物诱导的基因表达数据，预测有12种药物在使AD患者的基因表达恢复健康方面具有重要意义，其中包括9种缓解AD的药物。相比之下，如果药物预测是由在AD和健康对照中表达明显不同的基因进行的，则已知前12种重要药物中有4种可以缓解AD。转录组组织的多个水平的分析在建议与AD相关的共表达网络和发现药物方面很有用。如果药物预测是由在AD和健康对照中表达明显不同的基因进行的，则前12种重要药物中有4种可以缓解AD。转录组组织的多个水平的分析在建议与AD相关的共表达网络和发现药物方面很有用。如果药物预测是由在AD和健康对照中表达明显不同的基因进行的，则前12种重要药物中有4种可以缓解AD。转录组组织的多个水平的分析在建议与AD相关的共表达网络和发现药物方面很有用。